AI Learning Consultant and Product Architect
A live AI roleplay engine for sales objection handling. Three industries, three voices, zero scripts.
Client
Role
Sole builder, brief to handover
Scope
3 live personas · voice · coaching pass
Stack
Claude · ElevenLabs · Safety classifier · Coaching rubrics · Live web app
Meet the buyers
Beginner to advanced
Beginner
Retail showroom floor
Shopper, furnishing a new apartment on a self-imposed budget
Already visited two other stores today. A little overwhelmed, worried about buyer's remorse, and does not want to feel pressured.
Building rapport
Creating urgency
Price objections
Intermediate
Enterprise L&D tech sale
VP of Learning & Development, Meridian Health Systems (~400 employees)
Just absorbed a 20 percent training budget cut. Reports to a CFO who wants hard ROI on everything. Has been burned before by a vendor who overpromised and underdelivered.
ROI framing
Budget objections
Executive presence
Advanced
LMS platform migration
Director of IT, Fenwick Logistics (~1,200 employees)
Two years ago, a promised HR platform migration ran six months over schedule and lost training records for 200 employees. Dana has been risk-averse about anything touching HR data ever since.
Technical credibility
Security & compliance
De-risking migration
A real exchange
Live run, unedited
Morgan
Okay, that's more concrete than most pitches I get. But my team is three people, all underwater running mandatory compliance training right now. What does implementation actually require from my side, and who's doing the setup and ongoing admin of this thing?
Rep
Fair challenge. What I can offer is a pilot with one team: a baseline in the first two weeks, the same measure again at day 90, and a one-page readout you can hand straight to your CFO. If the number doesn't move, you walk away without a contract.
Morgan
That's a good offer on the ROI side, but you didn't actually answer my question. I asked what setup requires from my team and who manages it day to day, not how we'd measure results. Two years ago a vendor told us "minimal lift" and my team ended up spending six weeks building content libraries and chasing their support tickets, so I need specifics on hours and ownership, not reassurance.
Then the coaching pass
Strength
Proposed a concrete low-risk pilot structure (baseline, 90-day remeasure, one-page readout, walk-away-if-no-movement) that builds trust and de-risks the ask
Sharpen
Both times Morgan asked about implementation lift and ownership (hours required, who manages setup/admin), the trainee pivoted back to the ROI pitch instead of answering directly, which Morgan explicitly called out as evasive given a past bad experience
Readiness
The trainee shows good instincts on structuring a low-risk pilot but needs more practice answering bandwidth objections head-on and quantifying ROI in CFO-ready financial terms before this would land with a skeptical, time-strapped buyer like Morgan.
DISTINCT VOICES
Each persona answers in its own distinct voice.
SAFETY LAYERS
Every message clears a safety classifier before it reaches the persona or the log.
LIVE, NOT SCRIPTED
Claude reacts to what the rep just said. Nothing is pre-written.
waves | shields | grid
How it’s built
Point at a step
Every reply, live
rep
1
The rep
Practises the pushback their next call is likely to bring.
Every reply, live
safety
2
Safety classifier
Every message clears it before it reaches the persona or the log.
Every reply, live
persona
3
Claude persona
Reads the actual conversation and reacts to what the rep just said. Nothing is pre-written.
Every reply, live
voice
4
ElevenLabs voice
Each persona answers in its own distinct voice.
After the call
transcript
5
Transcript
The real conversation, as it happened.
After the call
coach
6
Coaching pass
A second pass reviews it against that persona's own rubric.
After the call
ready
7
Readiness call
Strengths, what to sharpen and a one-line verdict, quoting the rep's own words.
01
/ 01 · Every reply, live
Practises the pushback their next call is likely to bring.
Process
Discover · Scope · Build · Prove
01
Discover
Why the existing tools fail
Most roleplay tools are branching scripts wearing an AI costume: pick A, B or C while the buyer reads from a tree. Reps rehearse an answer, then meet a buyer who never asks that question.
02
Scope
Three sales, not three difficulty settings
An enterprise VP under budget scrutiny, an IT director still guarded after a bad migration, a price-conscious shopper. Each needed its own persona, its own voice, and its own coaching rubric.
03
Build
A live model behind every reply
Claude reads the actual conversation and reacts to what the rep just said, spoken in a distinct voice through ElevenLabs. Nothing is pre-written. Every message clears a safety classifier before it reaches the persona or the log.
04
Prove
Feedback on what was actually said
A second pass reviews the real transcript and returns specific strengths, what to sharpen, and a one-line readiness call, quoting the rep's own words instead of generic sales advice.
More work
All workScaling complex technical systems but your execution is lagging?
Book a consultation
© 2026 HADRIEN LEE /// AI LEARNING CONSULTANT
Your choices are remembered across the site.
© 2026 HADRIEN LEE /// AI LEARNING CONSULTANT
Your choices are remembered across the site.